New models often need a newer Ollama. If a fresh model fails with “unknown model architecture”, a pull errors out, or a feature from the release notes is missing, the fix is usually just to update Ollama.
The good news: updating never deletes your downloaded models, and on Windows and Mac it is one click. This guide shows how to check your version and update on every platform: Windows, macOS, Linux, Docker, winget and Homebrew. It also covers installing a specific older version, and fixing the problems people hit after an update.
- Check version:
ollama -v - Windows / Mac: click the Ollama tray or menu bar icon, then Restart to update. Or download the latest installer from ollama.com/download.
- Linux: re-run
curl -fsSL https://ollama.com/install.sh | sh - Docker:
docker pull ollama/ollama, then recreate the container with the same volume. - Your models are safe: they live in
~/.ollama/models(orOLLAMA_MODELS) and an update does not touch them.
New to Ollama? My Ollama guide covers install, models, commands and fixes in one place.
Why Update Ollama?
- New model support. Each new model family needs code changes. An old Ollama shows errors like “unknown model architecture” or refuses to pull the model.
- Speed and memory fixes. Releases regularly improve GPU offloading, Flash Attention and KV cache handling.
- New features. Things like thinking mode, tool calling improvements and app settings arrive through updates.
- Security fixes. Especially important if your Ollama server is reachable from other devices.
Ollama ships updates often, so checking every couple of weeks is a good habit if you try new models.
Check Your Current Version
ollama -v
You will see something like ollama version is 0.x.y. Compare it with the newest release on the Ollama GitHub releases page.
If you also see Warning: client version is ..., the command-line tool and the running server are different versions. That usually means an old server process is still running. Restarting Ollama fixes it (see Troubleshooting).
Update Ollama on Windows
Option 1: Built-in updater (easiest)
Ollama for Windows downloads updates in the background.
- Look for the Ollama icon in the system tray (bottom right, click the ^ arrow if hidden).
- If an update is ready, click the icon and choose Restart to update.
- Ollama closes, installs the new version and starts again. Run
ollama -vin a new terminal to confirm.
Option 2: Re-run the installer
- Download
OllamaSetup.exefrom ollama.com/download. - Quit Ollama from the tray icon.
- Run the installer. It installs over the old version for your user account, with no admin rights needed.
Option 3: winget
If you installed Ollama with winget, update it the same way:
winget upgrade Ollama.Ollama
Update logs are saved in %LOCALAPPDATA%\Ollama\upgrade.log, which is handy if an update fails. If you use AMD graphics, also see my AMD GPU for AI guide.
Update Ollama on macOS
Option 1: Menu bar
- Click the Ollama llama icon in the menu bar.
- Choose Restart to update when it appears.
Option 2: Download again
Quit Ollama, download the latest Mac app from ollama.com/download and drag it into Applications, replacing the old one.
Option 3: Homebrew
If you installed Ollama through Homebrew, update it with Homebrew instead, so the two installs do not get mixed up:
brew update
brew upgrade ollama
If you run the server with brew services, restart it afterwards with brew services restart ollama.
Update Ollama on Linux
Re-run the official install script. It replaces the binary and keeps your service, settings and models:
curl -fsSL https://ollama.com/install.sh | sh
The script restarts the ollama systemd service for you. Confirm with:
ollama -v
systemctl status ollama
Manual alternative (no script), straight from the official docs:
curl -fsSL https://ollama.com/download/ollama-linux-amd64.tar.zst | sudo tar x -C /usr
sudo systemctl restart ollama
Your custom settings in systemctl edit ollama.service (such as OLLAMA_HOST, OLLAMA_KEEP_ALIVE or OLLAMA_CONTEXT_LENGTH) live in an override file, so the update keeps them. To check the server logs after updating, run journalctl -e -u ollama.
Running Ollama inside WSL2? Use the Linux steps inside your WSL distro. My WSL2 local AI setup has more.
Update Ollama in Docker
A container never updates itself. Pull the new image and recreate the container with the same volume, so your models stay:
docker pull ollama/ollama
docker stop ollama
docker rm ollama
docker run -d --gpus=all -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
The -v ollama:/root/.ollama part is what keeps your models. Use exactly the same volume name or path as before. For AMD, use the ollama/ollama:rocm image instead.
With Docker Compose:
docker compose pull
docker compose up -d
Check the version inside the container with docker exec -it ollama ollama -v. More setups are in my Ollama Docker guide.
Install a Specific (Older) Version
Sometimes a new release has a bug that hits your setup, and you want to go back one version. Or you want a pre-release to try a new model early.
Linux: set OLLAMA_VERSION for the install script:
curl -fsSL https://ollama.com/install.sh | OLLAMA_VERSION=0.5.7 sh
Replace 0.5.7 with any version number from the releases page, including pre-releases.
Windows and Mac: download the installer for the version you want from the “Assets” section of that release on the GitHub releases page. The app will still offer the newest version later, so skip the restart prompt if you want to stay on the older one.
Docker: use a version tag instead of latest, for example ollama/ollama:0.5.7.
Do I Need to Re-Download Models?
No. Models are stored separately from the program:
| System | Default models folder |
|---|---|
| Windows | C:\Users\<you>\.ollama\models |
| macOS | ~/.ollama/models |
| Linux (service) | /usr/share/ollama/.ollama/models |
| Docker | Inside the volume mounted at /root/.ollama |
If you moved them with OLLAMA_MODELS, they stay wherever you put them. That setup is explained in my guide to moving Ollama models.
Occasionally a release improves a model’s template or settings in the library. Run ollama pull <model> again to get those updates. Only the changed parts download.
After Updating: Quick Checks
ollama -vshows the new version with no client warning.ollama liststill shows all your models.- Run a quick prompt, then check
ollama ps. The PROCESSOR column should still say100% GPUfor models that fit. - Open WebUI or other apps still connect at
http://localhost:11434.
Troubleshooting
“Warning: client version is …” after updating
The old server is still running. On Windows and Mac, quit Ollama from the tray or menu bar and start it again. On Linux, run sudo systemctl restart ollama. If it continues, you may have two installs (for example Homebrew plus the app). Check which ollama (Mac/Linux) or where ollama (Windows).
Ollama stopped using the GPU after an update
Restart the PC first, since driver and runtime changes sometimes need it. Then update your GPU driver and read the server log. Full steps are in my Ollama not using GPU fix.
“Restart to update” never appears
The background download may be blocked by a firewall or proxy, or auto-update may be turned off. Just download and run the latest installer manually.
Linux script fails with permission errors
The script needs sudo rights for /usr and systemd. Run it from a user with sudo access. On systems without systemd, the binary still updates but you must restart ollama serve yourself.
A model still says “unknown model architecture”
Brand-new architectures can take days to land in a stable release. Check the releases page for a pre-release that mentions the model, or run it in llama.cpp’s llama-server, which often gets support first.
My settings disappeared
Environment variables set in a terminal only last for that session. Save them as system environment variables (Windows), with launchctl setenv (Mac), or in a systemd override (Linux). See my keep_alive guide for the exact steps per platform.
FAQ
Does Ollama update automatically?
On Windows and macOS, the app downloads updates automatically and asks you to restart. On Linux and Docker, you update by hand.
Will updating delete my models or chats?
No. Models stay in the models folder. Chats live in your app (Open WebUI, the Ollama app and so on) and are not touched.
How often does Ollama release updates?
Often, usually several releases a month. Check the GitHub releases page for the changelog.
Can I turn off automatic updates?
The app downloads updates in the background, but it only installs them when you click Restart to update. If you need to stay on one version, check the app settings for an update option. Otherwise, simply skip the restart prompt, or run Ollama in Docker with a fixed version tag.
How do I completely uninstall instead?
See my guide to deleting models and uninstalling Ollama.
Sources
- Ollama FAQ: How can I upgrade Ollama? (auto-update on Windows/Mac, Linux install script)
- Ollama docs: Linux (updating, OLLAMA_VERSION, manual tarball, logs)
- Ollama docs: Windows (installer, upgrade.log, file locations)
- Ollama docs: Docker
- Ollama releases on GitHub



